WO2019033671A1 - Method and system for extracting source signal, and storage medium - Google Patents

Method and system for extracting source signal, and storage medium Download PDF

Info

Publication number
WO2019033671A1
WO2019033671A1 PCT/CN2017/117813 CN2017117813W WO2019033671A1 WO 2019033671 A1 WO2019033671 A1 WO 2019033671A1 CN 2017117813 W CN2017117813 W CN 2017117813W WO 2019033671 A1 WO2019033671 A1 WO 2019033671A1
Authority
WO
WIPO (PCT)
Prior art keywords
signal
input
input signal
signals
interfering
Prior art date
Application number
PCT/CN2017/117813
Other languages
French (fr)
Chinese (zh)
Inventor
张健钢
Original Assignee
音科有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 音科有限公司 filed Critical 音科有限公司
Priority to EP17921701.3A priority Critical patent/EP3672275A4/en
Publication of WO2019033671A1 publication Critical patent/WO2019033671A1/en

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R3/00—Circuits for transducers, loudspeakers or microphones
    • H04R3/005—Circuits for transducers, loudspeakers or microphones for combining the signals of two or more microphones
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R3/00—Circuits for transducers, loudspeakers or microphones
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G10L21/0272—Voice signal separating
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R25/00—Deaf-aid sets, i.e. electro-acoustic or electro-mechanical hearing aids; Electric tinnitus maskers providing an auditory perception
    • H04R25/40—Arrangements for obtaining a desired directivity characteristic
    • H04R25/407—Circuits for combining signals of a plurality of transducers
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R1/00—Details of transducers, loudspeakers or microphones
    • H04R1/20—Arrangements for obtaining desired frequency or directional characteristics
    • H04R1/32—Arrangements for obtaining desired frequency or directional characteristics for obtaining desired directional characteristic only
    • H04R1/40—Arrangements for obtaining desired frequency or directional characteristics for obtaining desired directional characteristic only by combining a number of identical transducers
    • H04R1/406—Arrangements for obtaining desired frequency or directional characteristics for obtaining desired directional characteristic only by combining a number of identical transducers microphones
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R2225/00—Details of deaf aids covered by H04R25/00, not provided for in any of its subgroups
    • H04R2225/43—Signal processing in hearing aids to enhance the speech intelligibility
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R2430/00—Signal processing covered by H04R, not provided for in its groups
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
    • H04R2430/00—Signal processing covered by H04R, not provided for in its groups
    • H04R2430/20—Processing of the output signals of the acoustic transducers of an array for obtaining a desired directivity characteristic
    • H04R2430/23—Direction finding using a sum-delay beam-former

Definitions

  • the present disclosure relates to the field of signal processing technologies, and in particular, to a method, system, and storage medium for extracting interference signals from a mixed signal.
  • the current separation technique mainly operates the hearing device by selectively adjusting the specific gravity of the signal, focusing on how to calculate the coefficient matrix more efficiently, or using a combination of a directional microphone and an omnidirectional microphone to enhance the clarity of the speech, but
  • the traditional independent component analysis (ICA) method can not achieve the desired effect, the removal effect of the interference signal is not ideal, and the accuracy of the ICA is destroyed.
  • ICA independent component analysis
  • the present disclosure creatively solves the technical problem of incomplete signal separation by simplifying the steps by performing time domain synchronization processing on the signals, and achieves an extremely high precision effect of removing interference signals.
  • An aspect of the present disclosure is to provide a method of removing a target interference signal from a multiple signal, the method comprising:
  • each of the input signals comprising both a useful signal and an interference signal
  • Another aspect of the present disclosure is to provide a system for removing a target interference signal from a multiple signal, the system comprising:
  • a memory storing computer readable instructions, when the processor executes the instructions, the processor can:
  • the present disclosure also provides a non-transitory computer storage medium, characterized in that the medium stores computer readable instructions, and when the processor executes the instruction, a target interference signal can be removed from the multiple signals.
  • Method comprising:
  • each of the input signals comprising both a useful signal and an interference signal
  • the application can eliminate or weaken the influence of the asynchronous effect and improve the signal source extraction performance, and can improve the perception of the target signal by continuously removing the interference signal even during the motion of the useful signal and the interference signal.
  • FIG. 1 is a flow chart of a method for removing a target interference signal from a multiple signal according to an embodiment of the present disclosure
  • FIG. 2 is a flowchart 1 of an operation flowchart of a synchronized input signal according to an embodiment of the present disclosure
  • FIG. 3 is a second flowchart of an operation flowchart of a synchronized input signal according to an embodiment of the present disclosure
  • FIG. 4 is a third flowchart of an operation flowchart of a synchronized input signal according to an embodiment of the present disclosure
  • FIG. 5 is a fourth flowchart of an operation flowchart of a synchronized input signal according to an embodiment of the present disclosure
  • FIG. 6 is a schematic structural diagram of a computer system for implementing an embodiment of the present disclosure
  • Figure 7 is a schematic view showing the position of different sound sources to different sensors
  • Figure 8 shows the signal delay of two sensors with a certain interval.
  • FIG. 1 is a flow diagram of a method 1000 of removing a target interference signal from an input signal in accordance with an embodiment of the present disclosure.
  • step 100 signals sent by m signal sources are first received using n signal receiving devices, and the mixed signal received by each signal receiving device is referred to as an input signal of the signal receiving device. Determining one of each input signal (a mixed signal of signals from m sources) or Signals from multiple sources are useful signals, others are interference signals.
  • the signal receiving device can be a sensor or a cloud platform.
  • the signal receiving device may also be an input data interface connected to the storage unit.
  • the storage unit stores signal data in advance, and the input data interface receives signal data from the storage unit.
  • each input signal may also include a plurality of interference signals that are not identical to each other. It can be understood that the interference signals in the input signal may also be the same, and the disclosure is not particularly limited thereto.
  • the electronic monitoring device typically includes at least two microphones, each of which can be used to receive a mixed signal composed of a sound source (useful signal) and an ambient background sound (interference signal). Since the microphone is usually placed in different positions, the useful signal and the interference signal are received by two or more microphones at different positions spaced apart from each other, so the ambient background sound received by the different microphones is in the time domain and/or amplitude. There is a difference between each other. As another example, in a studio recording and/or 360 audio recording scene, two or more microphones are used to measure the sound. Since the microphone is usually placed in a different position, the useful signal and the interference signal are two or more microphones.
  • a brain wave device generally includes at least two electrodes, each of which can receive a mixed signal including a brain wave source signal and an interference signal. Since the electrodes are usually placed at different positions, the useful signal and the disturbing noise are received by two or more electrodes at different positions spaced apart from each other, and the ambient noise received by the different electrodes is in time domain and/or amplitude with each other. There are differences.
  • the echo receiving device typically includes at least two sensors, each of which can be used to receive a mixed signal from the sound source and ambient noise.
  • the sensors are usually placed in different locations
  • the two or more sensors are received at different locations that are spaced apart from each other, and the ambient noise received by the different sensors is different in time domain and/or amplitude from each other.
  • the signals received by Mi, Mj should be composed of the following equations, each with a different amplitude a And different time delays are propagated to the sensors Mi and Mj.
  • M i a 1i S 1 (t 1 + ⁇ 1i ) + a 2i S 2 (t 2 + ⁇ 2i ) + ... + a ni S n (t n + ⁇ ni )
  • M j a 1j S 1 (t 1 + ⁇ 1j )+a 2j S 2 (t 2 + ⁇ 2j )+...+a nj S n (t n + ⁇ nj )
  • signals received by other sensors can be analogized by the same formula.
  • the energy of the sound will decrease inversely with increasing distance.
  • the following formula can be used to represent the sound signal received by the sensor:
  • the formula is expressed as follows. Note that in order to simplify the expression, the formula has simplified all constant terms to 1.
  • step 200 the coefficient matrix is decomposed to restore a portion of the mixed signal to a useful signal.
  • the independence of the mixed signal is improved by decomposing the coefficient matrix.
  • the independence of the mixed signal is maximized by decomposing the matrix of coefficients.
  • the coefficient matrix is decomposed by increasing the probability distribution of the mixed signal as far as possible from the normal distribution to improve the independence of the signal source. Specifically, taking the coefficient matrix parameter as the dependent variable, an objective function is set to calculate and measure whether the variable is close to the normal distribution, and the optimal parameter of the convergence of the objective function is calculated, and the decomposition parameter matrix is obtained.
  • step 200 selects the following function as the objective function for calculating and measuring whether the variable is close to a normal distribution:
  • E ⁇ represents the calculation of the expected value and y is the mixed signal.
  • the objective function value is 0, it means that the probability distribution of y is normally distributed.
  • Kurtosis can be replaced by other measures as a standard away from the normal distribution, and the disclosure does not have any specific limitation.
  • the objective function can be rewritten as the following formula:
  • the coefficient matrix parameter is used as the dependent variable, and the above formula is the objective function.
  • the Newton iterative method is used to find the optimal parameter of the objective function convergence, that is, the decomposition parameter matrix. The following is a brief list of specific calculation methods:
  • Step 300 synchronizing the input signals in the time domain. This step can be implemented by four different methods. In conjunction with Figures 2, 3, 4, and 5, this step 300 is specifically described as follows.
  • step 3101 is to intercept two or more discrete segments of the interference signal, and the duration of the discrete segments is controlled to be n milliseconds. If the signal is an audio signal, n needs to be greater than 0.98 milliseconds and less than 20.03 milliseconds. When the length n is controlled within this interval, the human can not hear the echo under the condition of ensuring the accuracy, so the real-time processing effect is the best, and the user's hearing effect is the best.
  • step 3101 continues to intercept discrete segments of each of the mixed signals in real time.
  • the method of this embodiment can process the signal in real time.
  • the target interference signal is determined by mode recognition, and the target interference signal is extracted.
  • the target interference signal is male
  • the pattern recognition will judge whether each discrete segment of the mixed signal is male, if it is male.
  • the segment is extracted to the next step, and if the interference signal is a female voice, the segment determined to be female voice is extracted to proceed to the next step.
  • the two sound sources are vocal or non-human.
  • the detection process of the interference signal of step 3101 can detect the presence of an interference signal from a low level to a high level within n milliseconds (ie, the interference signal begins a step signal response or from a high level to a low level, for example Set the voice of the man as the interference signal.
  • the interference signal begins a step signal response or from a high level to a low level, for example Set the voice of the man as the interference signal.
  • the man speaks he does not need to say the complete word. It only needs to detect the n milliseconds of the voice of the person speaking, that is, the sound is determined to be the interference signal. This method is greatly reduced.
  • the requirement for a complex signal (such as a sound signal) detection process thereby reducing the computational complexity and its cost.
  • step 3102 discrete time convolutions of the two detected interfering signal segments are calculated to obtain their time delay. Assuming that the two mixed signals are x, y, respectively, the correlation formula between the two signals is calculated as:
  • mx is the average of x
  • my is the average of y
  • d is the time delay.
  • the molecular part of this formula is the discrete time convolution.
  • the correlation formula is:
  • the d when the maximum value is generated in r(d) is the time delay.
  • the input signal is synchronized based on the acquired time delay d. For example, if the time delay of the first interfering signal detected from the first input signal f 1 (t) and the second interfering signal detected from the second input signal f 2 (t) is denoted by ⁇ , then the first The input signal f 1 (t) is delayed by ⁇ , i.e. corrected to f 1 (t- ⁇ ), thereby being synchronized with the second input signal f 2 (t).
  • the method can continuously update the iterative time delay and dynamically track the change of the interference signal when the signal source and the sensor move or move relatively differently.
  • step 3201 since a plurality of sensors are placed at different positions, the interference signal is received by two or more sensors at different positions spaced apart from each other.
  • the position of each interference signal relative to the sensor that is, the relative delay of each interference signal, is first calculated; then an interference signal is selected according to the relative delay of each interference signal.
  • selecting an interference signal can also be selected by the user in real time.
  • the distance between the signal source and the sensor 1 is d1
  • the distance from the sensor 2 is d2
  • the signal sampling rate is Fs
  • the signal propagation speed is v.
  • the formula for calculating the relative delay dir is as follows:
  • a time delay is extracted according to an interference signal region or a preset interference signal region selected by the user in real time.
  • step 3203 synchronization processing is performed according to the time delay extracted by 3202, which is the same as step 3103.
  • this embodiment selects interference signals from all relative delays.
  • all time delays are calculated and analyzed based on different signals (such as sound), sensor distance, and signal propagation speed.
  • step 3302 all possible time delays ⁇ 1 , ⁇ 2 , . . . , ⁇ n are extracted.
  • step 3303 the synchronization processing in step 3103 is repeated for each of the different time delays.
  • step 3401 the useful signal direction is selected or preset by the user in real time.
  • step 3402 the time delays for these directions are calculated.
  • step 3403 in the method for all signal directions obtained based on FIG. 4, in step 3403, the time delay of these useful signals is excluded in all possible directions, and step 3103 is repeated for each of the remaining different time delays. Synchronization in .
  • step 400 the synchronized input signal is separated into a channel containing the target interference signal and a channel containing no target interference signal.
  • step 400 is implemented by a multiplication of the synchronized signal matrix and the coefficient matrix determined in step 200.
  • the mixed signal is composed as follows:
  • two channels will be generated, one of which is In other words, the channel is composed of 96% S2 and 4% S1, and if the target interference signal is S1, the channel is selected and output. Therefore, in this case, the separation effect of synchronization is 96%.
  • the target interference signal is S2
  • the mixed signal of the synchronization S2 is multiplied by the coefficient matrix. Also select the appropriate channel output.
  • step 500 based on the two channels obtained in step 400, the relative energy of the signal can be selected according to different signals, and the channel with the relatively lower signal energy is selected as the output channel.
  • the method of calculating the signal energy can be the root mean square value of the signal.
  • the selection process is applied to the channel containing the target interference signal obtained in step 500 and the channel containing no target interference signal.
  • an output channel will be generated in different time delays.
  • the optimal channel is selected as the signal output based on feature detection (e.g., in the generated channel)
  • the channel with the least interference component of the target signal in the embodiment of Fig. 5, the optimal channel can be selected as the signal output according to the signal energy (e.g., the channel with the lowest energy of the target interference signal in the generated channel).
  • the step of further processing the separated useful signal and the interference signal may be further included, for example, performing frequency domain enhancement.
  • the separated useful audio signal can be personalized for frequency domain enhancement.
  • the present disclosure provides an apparatus that includes a processor, and a human interaction interface.
  • the device may also include, but is not limited to, a memory, a controller, an input and output module, Information receiving module.
  • the processor is configured to perform the above steps 100, 200, 3201-3203 (or 3401-3403), and 400, 500, and frequency domain enhancement (optional).
  • the user selects the area he wants to be the interference signal area in real time through the manual interaction interface.
  • the artificial interaction interface includes but is not limited to a voice receiving module, a sensor, and a video receiving module. Touch screen, keyboard, buttons, knobs, projection interface, virtual 3D interface.
  • the manner in which the user selects in real time through the manual interaction interface includes selecting a different identified area by voice instructions through different gestures or actions of the user.
  • the human interaction interface is a touch screen
  • the user can click on one of the areas, and the disclosure provides a user-controllable machine for removing interference signals, and the delay can be adjusted in real time.
  • steps 100-400 may occur in a different order than described in the drawings.
  • the order of the second embodiment of steps 100 and 300 ie, 3201-3203
  • any two of steps 100-400 may be performed in parallel or in reverse order, depending on the particular functionality involved.
  • step 200 is performed before step 300 by first calculating a coefficient matrix and then synchronizing the input signal in the time domain.
  • the advantage of this is that there is no need to recalculate the coefficient matrix based on different time delays. This can save a lot of calculations.
  • the result matrix can only be obtained by calculating the coefficient matrix once.
  • the present disclosure concludes by conducting a large number of experiments that the coefficient matrix calculated by the mixed signal after synchronization and the coefficient matrix calculated by the original mixed signal are almost the same. Therefore, this method saves a large amount of calculation without losing the accuracy of the coefficient matrix.
  • step 100 after receiving n input signals by using m signal receiving devices, determining whether to remove one or more signals of the plurality of signal receiving devices according to the determining condition Receiving an input signal received by the device.
  • the input signal is an acoustic signal and the signal receiving device is an acoustic signal receiving device (eg, a microphone).
  • the judgment condition is Fs*X/V ⁇ L/3 (where L is the length of the discrete signal intercepted, X is the distance of any two acoustic signal receiving devices, V is the signal propagation speed, Fs is the sampling rate), and two The acoustic signal received by one of the acoustic signal receiving devices is removed.
  • This embodiment reduces the amount of data that needs to be calculated while not affecting the accuracy of pattern recognition, improves computational efficiency, and reduces power consumption.
  • the signals include audio signals, image signals, electromagnetic signals, brain wave signals, electrical signals, radio wave signals, and other forms of signals that can be received by the sensor, and the disclosure is not particularly limited thereto.
  • the present disclosure can greatly improve the perception of the target signal while reducing the computational cost. Furthermore, the input signal of the present disclosure is synchronized in the time domain, so the method of the present disclosure minimizes frequency distortion.
  • FIG. 6 is a schematic diagram of the structure of a computer system 3000 suitable for implementing the above embodiments.
  • the computer system 3000 includes a central processing unit (CPU) 3001 that is executable in accordance with program instructions stored on an electrically programmable read only memory (EPROM) 3002 or random access memory (RAM) 3003. A variety of suitable operations and processes. The necessary programs and data for running the system 3000 can also be stored on the random access memory (RAM) 3003.
  • the central processing unit (CPU) 3001, the electrically programmable read only memory (EPROM) 3002, and the random access memory (RAM) 3003 are connected to one another via a bus 3004.
  • the bus 3004 is also connected to lose In/Out (I/O) interface 3005.
  • the bus 3004 is also coupled to a direct memory access interface 3006 to facilitate data exchange.
  • the input/output (I/O) interface 3005 is also connected to the following components: a removable data storage 3007, including a USB memory, a solid state hard disk, etc.; a wireless data transmission line 3008, including a local area network (LAN), a Bluetooth, a near field communication device (NFC); and a signal converter 3009 connected to the data input channel 3010 and the data output channel 3011.
  • a removable data storage 3007 including a USB memory, a solid state hard disk, etc.
  • LAN local area network
  • NFC near field communication device
  • signal converter 3009 connected to the data input channel 3010 and the data output channel 3011.
  • the process of the above flow chart may be implemented by an embedded computer system without a keyboard, mouse, and hard disk similar to the computer system 3000.
  • the processor may be a cloud processor, and the memory may be a cloud memory.
  • the process involved in the above flowchart may be implemented by a computer software program.
  • embodiments of the present disclosure provide a computer program product comprising a computer program stored in a tangible machine readable medium, the program comprising program code for executing the method shown in the flowchart.
  • the computer program can be downloaded and installed via wireless data transmission line 3008 and/or installed from removable medium 3007.
  • each block in the flowcharts and block diagrams represents a module, a program segment, or a code unit.
  • the module, program segment or code unit contains one or more executable instructions for implementing the specified logical function.
  • the functions represented by the modules may occur in a different order than that described in the drawings. For example, in practical applications, connected according to the specific functions involved. The operations shown in the two blocks may be performed in parallel or in reverse order.
  • each block or combination of the flowcharts and/or block diagrams can be implemented by a dedicated, hardware-based system that can perform a particular function or operation, or through dedicated software and computer instructions. The combination is implemented.
  • the unit or module involved in the embodiments of the present disclosure may be implemented by software or hardware, and the unit or module may be installed in a processor.
  • the name of the unit or module should not limit the unit or module itself.
  • the present disclosure also provides a computer readable storage medium.
  • the computer readable storage medium may be a computer readable storage medium installed in an instrument device to which the above embodiments are applied, or may be a separate computer readable storage medium that is not assembled in an instrument device.
  • the computer readable storage medium stores one or more programs that are executed by one or more processors to implement the method of sorting a target signal from a noise signal of the present disclosure.

Landscapes

  • Engineering & Computer Science (AREA)
  • Signal Processing (AREA)
  • Acoustics & Sound (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Otolaryngology (AREA)
  • General Health & Medical Sciences (AREA)
  • Neurosurgery (AREA)
  • Computational Linguistics (AREA)
  • Quality & Reliability (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Circuit For Audible Band Transducer (AREA)

Abstract

Provided are a method and system for continuously extracting target interference signals from selected signals, and a storage medium. The method comprises: collecting a two-channel or multi-channel input signal, each channel of the input signal containing a target interference signal; increasing independence of the input signal; calculating to obtain a resulting coefficient matrix after the independence of the input signal has been increased; synchronizing each pair or each group of input signals; separating the synchronized input signals into the target interference signal and a desired signal; and intelligently selecting an output signal.

Description

一种源信号提取方法、***和存储介质Source signal extraction method, system and storage medium 技术领域Technical field
本公开涉及信号处理技术领域，尤其涉及一种用于从混合信号中提取干扰信号的方法、***和存储介质。The present disclosure relates to the field of signal processing technologies, and in particular, to a method, system, and storage medium for extracting interference signals from a mixed signal.
背景技术Background technique
在当前信号处理和大数据领域，观察信号的测量经常会受到无用信号的干扰，因此，如何提高所测观察信号的信噪比是一个巨大挑战。同样的问题也出现在录音(如摄影棚录音、助听器、360音频器)、生物医学应用(如脑电波记录、脑成像)以及遥感(如雷达信号、回声定位)等领域。消除这类干扰信号最常用的方法是利用模拟或数字形式的滤波器。但是，有用信号和干扰信号经常共用一个频段，滤波器很难将它们分离开来。In the current signal processing and big data fields, the measurement of observation signals is often interfered by unwanted signals. Therefore, how to improve the signal-to-noise ratio of the observed signals is a huge challenge. The same problems occur in recording (such as studio recording, hearing aids, 360 audio), biomedical applications (such as brainwave recording, brain imaging), and remote sensing (such as radar signals, echolocation). The most common way to eliminate such interfering signals is to use filters in either analog or digital form. However, useful and interfering signals often share a single band, and it is difficult for the filter to separate them.
当前分离技术主要是通过选择性调节信号比重的方式来操作听力装置，重点聚焦在如何更有效地计算系数矩阵，或者采用一个方向性麦克风和一个全方向麦克风的组合来增强语音的清晰度，但传统的独立成分分析方法(ICA)无法达到理想的效果，对干扰信号的去除效果不理想，而且破坏了ICA的准确性。The current separation technique mainly operates the hearing device by selectively adjusting the specific gravity of the signal, focusing on how to calculate the coefficient matrix more efficiently, or using a combination of a directional microphone and an omnidirectional microphone to enhance the clarity of the speech, but The traditional independent component analysis (ICA) method can not achieve the desired effect, the removal effect of the interference signal is not ideal, and the accuracy of the ICA is destroyed.
因此，当前急需一种能有效地将有用信号和干扰信号分离出来的技术。 Therefore, there is an urgent need for a technique that can effectively separate useful and interference signals.
公开内容Public content
针对上述技术存在的问题，本公开创造性地通过对信号做时域同步化处理等方法，在简化步骤的同时解决了信号分离不彻底的技术问题，达到了极高精度的去除干扰信号的效果。In view of the problems of the above techniques, the present disclosure creatively solves the technical problem of incomplete signal separation by simplifying the steps by performing time domain synchronization processing on the signals, and achieves an extremely high precision effect of removing interference signals.
本公开的一方面是提供一种从多重信号中去除目标干扰信号的方法，该方法包括：An aspect of the present disclosure is to provide a method of removing a target interference signal from a multiple signal, the method comprising:
接收一组输入信号，该组输入信号中的每条输入信号既包含有用信号也包含干扰信号；Receiving a set of input signals, each of the input signals comprising both a useful signal and an interference signal;
提高所述输入信号的独立性；Increasing the independence of the input signal;
计算提升所述输入信号的独立性所得的系数矩阵；Calculating a coefficient matrix obtained by increasing the independence of the input signal;
同步所述输入信号；Synchronizing the input signal;
将同步后的输入信号分离为含有目标干扰信号的频道和不含目标干扰信号的频道；Separating the synchronized input signal into a channel containing the target interference signal and a channel containing no target interference signal;
智能选择合适的、不含目标干扰信号的频道作为信号输出。Intelligently select the appropriate channel without the target interference signal as the signal output.
本公开的另一方面是提供一种从多重信号中去除目标干扰信号的***，该***包括：Another aspect of the present disclosure is to provide a system for removing a target interference signal from a multiple signal, the system comprising:
一套用于输入两路或多路信号的输入设备；a set of input devices for inputting two or more signals;
处理器；以及Processor;
存储计算机可读指令的存储器，当所述处理器执行所述指令时，该处理器可进行：A memory storing computer readable instructions, when the processor executes the instructions, the processor can:
提升所述输入信号的独立性； Enhancing the independence of the input signal;
计算在输入通道提升所述输入信号的独立性所得的系数矩阵；Calculating a coefficient matrix obtained by increasing the independence of the input signal in the input channel;
同步所述输入信号；Synchronizing the input signal;
将同步后的输入信号分离为含有目标干扰信号的频道和不含目标干扰信号的频道；Separating the synchronized input signal into a channel containing the target interference signal and a channel containing no target interference signal;
智能选择合适的、不含目标干扰信号的频道作为信号输出。Intelligently select the appropriate channel without the target interference signal as the signal output.
另一方面，本公开还提供一种非临时性计算机存储介质，其特征在于，该介质存储有计算机可读指令，当处理器执行该指令时，可实现一种从多重信号中去除目标干扰信号的方法，所述方法包括：In another aspect, the present disclosure also provides a non-transitory computer storage medium, characterized in that the medium stores computer readable instructions, and when the processor executes the instruction, a target interference signal can be removed from the multiple signals. Method, the method comprising:
接收一组输入信号，该组输入信号中的每条输入信号既包含有用信号也包含干扰信号；Receiving a set of input signals, each of the input signals comprising both a useful signal and an interference signal;
提升所述输入信号的独立性；Enhancing the independence of the input signal;
计算提升所述输入信号的独立性所得的系数矩阵；Calculating a coefficient matrix obtained by increasing the independence of the input signal;
同步所述输入信号；Synchronizing the input signal;
将同步后的输入信号分离为含有目标干扰信号的频道和不含目标干扰信号的频道；Separating the synchronized input signal into a channel containing the target interference signal and a channel containing no target interference signal;
智能选择合适的、不含目标干扰信号的频道作为信号输出。Intelligently select the appropriate channel without the target interference signal as the signal output.
本申请能消除或减弱异步化影响，提高信号源提取性能，即便在有用信号和干扰信号的运动过程中，也能通过持续去除干扰信号从而改善目标信号的感知度。The application can eliminate or weaken the influence of the asynchronous effect and improve the signal source extraction performance, and can improve the perception of the target signal by continuously removing the interference signal even during the motion of the useful signal and the interference signal.
附图说明 DRAWINGS
下面将参照附图对本公开的实施方式进行示例而非限制性的描述。附图是示范性的且不受图中表现出来的比例尺的限制。不同附图中相同的或相似的元件采用相同的符号标记。The embodiments of the present disclosure are exemplified and not described in the following with reference to the accompanying drawings. The drawings are exemplary and are not limited by the scales shown in the figures. Identical or similar elements in different figures are labeled with the same symbols.
图1是本公开实施例的一种从多重信号中去除目标干扰信号的方法的流程图；1 is a flow chart of a method for removing a target interference signal from a multiple signal according to an embodiment of the present disclosure;
图2是本公开实施例的同步化输入信号的操作流程图方法一；2 is a flowchart 1 of an operation flowchart of a synchronized input signal according to an embodiment of the present disclosure;
图3是本公开实施例的同步化输入信号的操作流程图方法二；3 is a second flowchart of an operation flowchart of a synchronized input signal according to an embodiment of the present disclosure;
图4是本公开实施例的同步化输入信号的操作流程图方法三；4 is a third flowchart of an operation flowchart of a synchronized input signal according to an embodiment of the present disclosure;
图5是本公开实施例的同步化输入信号的操作流程图方法四；5 is a fourth flowchart of an operation flowchart of a synchronized input signal according to an embodiment of the present disclosure;
图6是本公开的一种用于实现本公开实施例的计算机***结构示意图；6 is a schematic structural diagram of a computer system for implementing an embodiment of the present disclosure;
图7为不同声源至不同传感器的位置示意图；Figure 7 is a schematic view showing the position of different sound sources to different sensors;
图8示两个有一定间隔的传感器的信号延迟。Figure 8 shows the signal delay of two sensors with a certain interval.
具体实施方式Detailed ways
下文将结合附图详细描述本公开的具体实施例。Specific embodiments of the present disclosure will be described in detail below with reference to the drawings.
图1是本公开实施例的一种从输入信号中去除目标干扰信号的方法1000的流程图。1 is a flow diagram of a method 1000 of removing a target interference signal from an input signal in accordance with an embodiment of the present disclosure.
在步骤100中，首先使用n个信号接收装置接收m个信号源发出的信号，每个信号接收装置接收的混合信号称为该信号接收装置的输入信号。决定每条输入信号(均为m个信号源发出的信号的混合信号)中的一个或 多个信号源发出的信号为有用信号，其他为干扰信号。所述信号接收装置可以是传感器或云平台。信号接收装置也可以是输入数据接口，与存储单元相连接，存储单元预先存储有信号数据，输入数据接口从存储单元里接收信号数据。此外，每条输入信号还可以包括多种彼此并不相同的干扰信号。可以理解的是，输入信号中的这些干扰信号也可以是相同的，本公开对此并无特殊限制。例如，在电子监听装置的场景中，电子监听装置通常包含至少两个麦克风，每个麦克风都可用于接收由发声源(有用信号)和环境背景音效(干扰信号)构成的混合信号。由于麦克风通常安放在不同的位置，有用信号和干扰信号被两个或更多的麦克风在相互有间隔的不同位置接收，所以由不同的麦克风接收到的环境背景音效在时域和/或幅度上彼此是有差异的。再比如，在摄影棚录制和/或360音频录制场景中，用两个或更多的麦克风测量音效，由于麦克风通常安放在不同的位置，因此有用信号和干扰信号被两个或更多的麦克风在相互有间隔的不同位置接收，由不同的麦克风接收到的环境背景音在时域和/或幅度上彼此是有差异的。又比如在脑-机接口装置场景中,脑电波设备通常包含至少两个电极，每个电极都可接收包含由脑波源信号和干扰信号构成的混合信号。由于电极通常安放在不同的位置，因此有用信号和干扰噪音被两个或更多的电极在相互有间隔的不同位置接收，由不同的电极接收到的环境噪音在时域和/或幅度上彼此是有差异的。同样的，在水下回波检测场景中，回波接收装置通常包括至少两个传感器，每个传感器都可用于接收来自声源和环境噪音的混合信号。由于传感器通常安放在不同的位置，因此有用信号和干扰信号 被两个或更多的传感器在相互有间隔的不同位置接收，由不同的传感器接收到的环境噪音在时域和/或幅度上彼此是有差异的。假设有两个不同的传感器Mi，Mj和多个不同的信号源S1，S2,…Sn,则Mi,Mj所接收到的信号应由如下公式构成，每个不同的信号源以不同的幅度a及不同的时间延迟传播到传感器Mi和Mj。In step 100, signals sent by m signal sources are first received using n signal receiving devices, and the mixed signal received by each signal receiving device is referred to as an input signal of the signal receiving device. Determining one of each input signal (a mixed signal of signals from m sources) or Signals from multiple sources are useful signals, others are interference signals. The signal receiving device can be a sensor or a cloud platform. The signal receiving device may also be an input data interface connected to the storage unit. The storage unit stores signal data in advance, and the input data interface receives signal data from the storage unit. In addition, each input signal may also include a plurality of interference signals that are not identical to each other. It can be understood that the interference signals in the input signal may also be the same, and the disclosure is not particularly limited thereto. For example, in the context of an electronic monitoring device, the electronic monitoring device typically includes at least two microphones, each of which can be used to receive a mixed signal composed of a sound source (useful signal) and an ambient background sound (interference signal). Since the microphone is usually placed in different positions, the useful signal and the interference signal are received by two or more microphones at different positions spaced apart from each other, so the ambient background sound received by the different microphones is in the time domain and/or amplitude. There is a difference between each other. As another example, in a studio recording and/or 360 audio recording scene, two or more microphones are used to measure the sound. Since the microphone is usually placed in a different position, the useful signal and the interference signal are two or more microphones. Received at different locations that are spaced apart from each other, the ambient background sounds received by the different microphones differ from each other in time domain and/or amplitude. For example, in a brain-computer interface device scenario, a brain wave device generally includes at least two electrodes, each of which can receive a mixed signal including a brain wave source signal and an interference signal. Since the electrodes are usually placed at different positions, the useful signal and the disturbing noise are received by two or more electrodes at different positions spaced apart from each other, and the ambient noise received by the different electrodes is in time domain and/or amplitude with each other. There are differences. Similarly, in underwater echo detection scenarios, the echo receiving device typically includes at least two sensors, each of which can be used to receive a mixed signal from the sound source and ambient noise. Useful signals and interference signals because the sensors are usually placed in different locations The two or more sensors are received at different locations that are spaced apart from each other, and the ambient noise received by the different sensors is different in time domain and/or amplitude from each other. Assuming there are two different sensors Mi, Mj and a number of different signal sources S1, S2, ... Sn, then the signals received by Mi, Mj should be composed of the following equations, each with a different amplitude a And different time delays are propagated to the sensors Mi and Mj.
Mi＝a1iS1(t1+τ1i)+a2iS2(t2+τ2i)+…+aniSn(tn+τni)M i = a 1i S 1 (t 1 + τ 1i ) + a 2i S 2 (t 2 + τ 2i ) + ... + a ni S n (t n + τ ni )
Mj＝a1jS1(t1+τ1j)+a2jS2(t2+τ2j)+…+anjSn(tn+τnj)M j = a 1j S 1 (t 1 +τ 1j )+a 2j S 2 (t 2 +τ 2j )+...+a nj S n (t n +τ nj )
同理，其他的传感器接收到的信号亦可以相同公式类推。Similarly, signals received by other sensors can be analogized by the same formula.
为简化表述，如图7所述，为两个传感器和两个信号源在二维空间内的位置。请注意，该图只为简化解释于二维平面内表述，所有位置可拓展至一维、三维或者更高维度的表述。为了更加简化表述，以声学信号为例，假设S1，S2为两个声源，M1，M2为麦克风。假设声音传播速度为v,同时假设该传感器的采样率为Fs。因此声源到传感器的传播时间可以用如下公式表示：To simplify the description, as shown in Figure 7, the position of the two sensors and the two signal sources in a two-dimensional space. Please note that this figure is only for simplified interpretation in a two-dimensional plane, and all positions can be extended to one-dimensional, three-dimensional or higher dimensional representations. To simplify the expression, take the acoustic signal as an example. Suppose S1 and S2 are two sound sources, and M1 and M2 are microphones. Assume that the sound propagation speed is v, and that the sampling rate of the sensor is assumed to be Fs. Therefore, the propagation time from the sound source to the sensor can be expressed by the following formula:
tij＝Fs*dis{Si，Mj}/v   (1)t ij =Fs*dis{S i ,M j }/v (1)
在一个随机实施例里，v＝34029cm/s,Fs＝44.1kHZ。In a random embodiment, v = 34029 cm/s and Fs = 44.1 kHZ.
理想情况下，声音的能量会随着距离的增加成反比减少，则可用下列公式表示传感器接收到的声音信号：Ideally, the energy of the sound will decrease inversely with increasing distance. The following formula can be used to represent the sound signal received by the sensor:
Figure PCTCN2017117813-appb-000001
Figure PCTCN2017117813-appb-000001
具体到图7所述，该公式表述如下，请注意为了简化表述，该公式已将所有常数项简化为1。Specifically, as shown in FIG. 7, the formula is expressed as follows. Note that in order to simplify the expression, the formula has simplified all constant terms to 1.
Figure PCTCN2017117813-appb-000002
Figure PCTCN2017117813-appb-000002
由于实际情况中，S1，S2及公式中右边所示的系数矩阵是未知的；公式左半边的M1real和M1real是M1和M2麦克风所接收到的混合信号。下一步，步骤200，分解系数矩阵，将混合信号的一部分还原为有用信号。Since the actual situation, S1, S2 and the coefficient matrix shown on the right side of the formula are unknown; M 1real and M 1real in the left half of the formula are the mixed signals received by the M1 and M2 microphones. Next, in step 200, the coefficient matrix is decomposed to restore a portion of the mixed signal to a useful signal.
在步骤200中，通过分解系数矩阵来提高混合信号的独立性。优选的，通过分解系数矩阵来最大化混合信号的独立性。本实施例的前提假设为：每一个信号源都是相互独立的，然后根据中心极限定理的统计概率理论(即多个独立变量之和的概率统计分布会比每一个独立变量的概率统计分布更趋向于正态分布)，判断实施例混合信号的概率统计分布会比每一个信号源的概率分布更趋向于正态分布。由此，正态在本实施例中，通过尽可能将混合信号的概率分布统计远离正态分布来分解系数矩阵以提高信号源的独立性。具体的，以系数矩阵参数为因变量，设置一目标函数来计算并衡量变量是否接近正态分布，计算得到目标函数收敛的最佳参数，即得出分解参数矩阵。In step 200, the independence of the mixed signal is improved by decomposing the coefficient matrix. Preferably, the independence of the mixed signal is maximized by decomposing the matrix of coefficients. The premise of this embodiment assumes that each signal source is independent of each other and then according to the statistical probability theory of the central limit theorem (that is, the probability distribution of the sum of multiple independent variables will be more than the probability distribution of each independent variable. It tends to be normally distributed. It is judged that the probability statistical distribution of the mixed signal of the embodiment tends to be more normally distributed than the probability distribution of each signal source. Thus, in the present embodiment, the coefficient matrix is decomposed by increasing the probability distribution of the mixed signal as far as possible from the normal distribution to improve the independence of the signal source. Specifically, taking the coefficient matrix parameter as the dependent variable, an objective function is set to calculate and measure whether the variable is close to the normal distribution, and the optimal parameter of the convergence of the objective function is calculated, and the decomposition parameter matrix is obtained.
例如：步骤200选取下列函数作为计算并衡量变量是否接近正态分布的目标函数：For example, step 200 selects the following function as the objective function for calculating and measuring whether the variable is close to a normal distribution:
kurt(y)＝E{y4}-3(E{y2})2    (4) Kurt(y)=E{y 4 }-3(E{y 2 }) 2 (4)
E{}代表了计算期望值，y为混合信号。当目标函数值为0时，即表明，y的概率分布呈正态分布。当然也可以由其他衡量方式来替代Kurtosis作为远离正态分布的标准,本公开对此并没有特定的限制。对于此公式，目标函数可改写为以下公式：E{} represents the calculation of the expected value and y is the mixed signal. When the objective function value is 0, it means that the probability distribution of y is normally distributed. Of course, Kurtosis can be replaced by other measures as a standard away from the normal distribution, and the disclosure does not have any specific limitation. For this formula, the objective function can be rewritten as the following formula:
J(y)∝[E{G(y)}-E{G(v)}]2      (5)J(y)∝[E{G(y)}-E{G(v)}] 2 (5)
因此以系数矩阵参数为因变量，以上公式为目标函数，通过牛顿迭代方法寻找到目标函数收敛的最佳参数，即分解参数矩阵。下面简要列出具体计算方法：Therefore, the coefficient matrix parameter is used as the dependent variable, and the above formula is the objective function. The Newton iterative method is used to find the optimal parameter of the objective function convergence, that is, the decomposition parameter matrix. The following is a brief list of specific calculation methods:
1.Choose an initial(e.g.random)weight vector w.1.Choose an initial(e.g.random)weight vector w.
2.Let w+＝E{xg(wTx)}-E{g′(wTx)}w2.Let w + =E{xg(w T x)}-E{g'(w T x)}w
3.Let w＝w+/||w+||3.Let w=w + /||w + ||
4.If not converged，go back to 2.4.If not converged, go back to 2.
其中g为G的导函数。Where g is the derivative of G.
步骤300，在时域上同步所述输入信号。该步骤可由四种不同的方法实现，结合附图2、3、4、5，本步骤300具体说明如下。 Step 300, synchronizing the input signals in the time domain. This step can be implemented by four different methods. In conjunction with Figures 2, 3, 4, and 5, this step 300 is specifically described as follows.
如图2所示，步骤3101是截取干扰信号的两个或更多的离散片段，离散片段的时长控制在n毫秒。若信号为音频信号，n需要大于0.98毫秒，小于20.03毫秒。当时长n控制在这个区间内时，在保证精确度的情况下使得人类听不到回声，因此实时处理效果最好，用户听觉效果最佳。 As shown in FIG. 2, step 3101 is to intercept two or more discrete segments of the interference signal, and the duration of the discrete segments is controlled to be n milliseconds. If the signal is an audio signal, n needs to be greater than 0.98 milliseconds and less than 20.03 milliseconds. When the length n is controlled within this interval, the human can not hear the echo under the condition of ensuring the accuracy, so the real-time processing effect is the best, and the user's hearing effect is the best.
优选的，步骤3101实时持续截取混合信号的每个的离散片段。此实施例的方法可以实时对信号处理。Preferably, step 3101 continues to intercept discrete segments of each of the mixed signals in real time. The method of this embodiment can process the signal in real time.
然后，针对于每个离散片段区间的混合信号，通过模式识别的方式判断该离散片段是否为目标干扰信号，并提取目标干扰信号。例如，在声学案例中，有两个声源分别为男人和女人，假设目标干扰信号为男声，则该模式识别会对混合信号的每个n毫秒的离散片段做判断是否为男声，若为男声则将该片段提取进行下一步骤，若干扰信号为女声，则将判断为女声的片段提取进行下一步骤。再例如，两个声源分别为人声或非人声。本领域一般技术人员应该明白其他合理的方式也是可行的。Then, for the mixed signal of each discrete segment interval, whether the discrete segment is the target interference signal is determined by mode recognition, and the target interference signal is extracted. For example, in the acoustic case, there are two sound sources for men and women. If the target interference signal is male, the pattern recognition will judge whether each discrete segment of the mixed signal is male, if it is male. Then, the segment is extracted to the next step, and if the interference signal is a female voice, the segment determined to be female voice is extracted to proceed to the next step. For another example, the two sound sources are vocal or non-human. Those of ordinary skill in the art will appreciate that other reasonable ways are also possible.
步骤3101的干扰信号的检测过程可以通过在n毫秒内检测到有干扰信号从低电平到高电平(即，干扰信号开始阶跃信号响应或者从高电平到低电平，举个例子，设定男人发出的声音为干扰信号，男人说话时候不需要说完整个词，只需要探测到这个人说话的嗓音出现的n毫秒，即确定其声音为干扰信号。该方法极大的降低了对复杂的信号(例如声音信号)检测过程的要求，因而降低了计算的复杂度及其成本。The detection process of the interference signal of step 3101 can detect the presence of an interference signal from a low level to a high level within n milliseconds (ie, the interference signal begins a step signal response or from a high level to a low level, for example Set the voice of the man as the interference signal. When the man speaks, he does not need to say the complete word. It only needs to detect the n milliseconds of the voice of the person speaking, that is, the sound is determined to be the interference signal. This method is greatly reduced. The requirement for a complex signal (such as a sound signal) detection process, thereby reducing the computational complexity and its cost.
在步骤3102中，计算两个检测到的干扰信号片段的离散时间卷积以获取它们的时间延迟。假设两个混合信号分别为x,y，则计算该两个信号之间的相关性公式为：In step 3102, discrete time convolutions of the two detected interfering signal segments are calculated to obtain their time delay. Assuming that the two mixed signals are x, y, respectively, the correlation formula between the two signals is calculated as:
Figure PCTCN2017117813-appb-000003
Figure PCTCN2017117813-appb-000003
其中，mx是x的平均值，my是y的平均值，d为时间延迟，此公式的分子部分即为离散时间卷积。Where mx is the average of x, my is the average of y, and d is the time delay. The molecular part of this formula is the discrete time convolution.
通过不同d，即不同时间延迟的选取，其相关性公式为：Through different d, that is, the selection of different time delays, the correlation formula is:
Figure PCTCN2017117813-appb-000004
Figure PCTCN2017117813-appb-000004
基于此，在r(d)中选取最大值产生时的那个d即为时间延迟。Based on this, the d when the maximum value is generated in r(d) is the time delay.
在步骤3103中，基于获取的时间延迟d同步化处理所述输入信号。例如，如果从第一输入信号f1(t)中检测到的第一干扰信号和从第二输入信号f2(t)中检测到的第二干扰信号的时间延迟记作δ，则第一输入信号f1(t)被延迟时间δ，即修正成f1(t-δ)，由此与第二输入信号f2(t)同步。在另一实施例中，如果从第一输入信号f1(t)中检测到的干扰信号和从第二输入信号f2(t)中检测到的干扰信号的时间延迟记作-δ，则第一输入信号f1(t)被同步为f1(t+δ)。由于此实施例中持续地实时监测干扰信号片段，此方法可以在信号源和传感器不同移动或者相对移动时，不停地更新迭代时间延迟，动态地跟踪干扰信号的变化。In step 3103, the input signal is synchronized based on the acquired time delay d. For example, if the time delay of the first interfering signal detected from the first input signal f 1 (t) and the second interfering signal detected from the second input signal f 2 (t) is denoted by δ, then the first The input signal f 1 (t) is delayed by δ, i.e. corrected to f 1 (t-δ), thereby being synchronized with the second input signal f 2 (t). In another embodiment, if the time delay of the interference signal detected from the first input signal f 1 (t) and the interference signal detected from the second input signal f 2 (t) is denoted by -δ, then The first input signal f 1 (t) is synchronized to f 1 (t+δ). Since the interference signal segment is continuously monitored in real time in this embodiment, the method can continuously update the iterative time delay and dynamically track the change of the interference signal when the signal source and the sensor move or move relatively differently.
参见图3，步骤3201中，由于多个传感器被安放在不同的位置，因此干扰信号被两个或更多的传感器在相互有间隔的不同位置接收。此实施例是先行计算各个干扰信号相对于传感器的位置，即各个干扰信号的相对延迟；然后根据各个干扰信号的相对延迟选取一个干扰信号。其中，选取一个干扰信号也可以由用户实时选择。 Referring to Fig. 3, in step 3201, since a plurality of sensors are placed at different positions, the interference signal is received by two or more sensors at different positions spaced apart from each other. In this embodiment, the position of each interference signal relative to the sensor, that is, the relative delay of each interference signal, is first calculated; then an interference signal is selected according to the relative delay of each interference signal. Among them, selecting an interference signal can also be selected by the user in real time.
优选的，假设信号源到传感器1之间的距离为d1,到传感器2之间的距离为d2,信号采样率为Fs，信号传播速度为v。其相对延迟dir的计算公式如下：Preferably, it is assumed that the distance between the signal source and the sensor 1 is d1, the distance from the sensor 2 is d2, the signal sampling rate is Fs, and the signal propagation speed is v. The formula for calculating the relative delay dir is as follows:
dir＝Fs*(d1-d2)/v  (8)Dir=Fs*(d1-d2)/v (8)
假设传感器之间的距离为d，最大的方向Max(dir)由如下公式计算得到：Assuming that the distance between the sensors is d, the maximum direction Max(dir) is calculated by the following formula:
Max(dir)＝Fs*d/v  (9)Max(dir)=Fs*d/v (9)
若上述结果不是整数，采取四舍五入法取得整数。则所有的方向有：-Max(dir),…,-1,0,1,…,Max(dir)。If the above result is not an integer, rounding is used to obtain an integer. Then all directions are: -Max(dir),...,-1,0,1,...,Max(dir).
详见图8中距离描述的方向，假设采样率(Fs)为48kHZ，两个传感器(在本例子中为声学信号，所以此传感器为麦克风)之间的距离(d)为2.47cm,声音在空气中传播的速度(v)为340m/s，所以最大的延时为3。则可以分为7块区域，分别为延时为-3、-2、-1、0、1、2、3。在图8例子中，如预设干扰信号来自延时为-3的区域，则将延时固定为-3。See Figure 8 for the direction described by the distance. Assuming a sampling rate (Fs) of 48 kHz, the distance between the two sensors (in this case, the acoustic signal, so the sensor is a microphone) is 2.47 cm. The speed of propagation in the air (v) is 340 m/s, so the maximum delay is 3. It can be divided into 7 areas, the delay is -3, -2, -1, 0, 1, 2, 3. In the example of Fig. 8, if the preset interference signal comes from an area with a delay of -3, the delay is fixed to -3.
参见图3，步骤3202中，根据用户实时选择的干扰信号区域或者预设的干扰信号区域，提取时间延迟。Referring to FIG. 3, in step 3202, a time delay is extracted according to an interference signal region or a preset interference signal region selected by the user in real time.
参见图3，步骤3203中，根据3202提取的时间延迟做同步化处理，同3103步骤。Referring to FIG. 3, in step 3203, synchronization processing is performed according to the time delay extracted by 3202, which is the same as step 3103.
参见图4，此实施例是选择来自所有相对延迟的干扰信号。在步骤3301中，根据不同的信号(如声音)，传感器距离，信号传播速度进行分析计算所有的时间延迟。Referring to Figure 4, this embodiment selects interference signals from all relative delays. In step 3301, all time delays are calculated and analyzed based on different signals (such as sound), sensor distance, and signal propagation speed.
参见图4，在步骤3302中，提取所有可能的时间延迟τ1,τ2,…,τn, Referring to FIG. 4, in step 3302, all possible time delays τ 1 , τ 2 , . . . , τ n are extracted.
参见图4，在步骤3303中，对每一个不同的时间延迟分别重复做步骤3103中的同步化处理。Referring to FIG. 4, in step 3303, the synchronization processing in step 3103 is repeated for each of the different time delays.
参见图5，在步骤3401中，通过用户实时选择或者预设的有用信号方向。Referring to FIG. 5, in step 3401, the useful signal direction is selected or preset by the user in real time.
参见图5，在步骤3402中，计算这些方向的时间延迟。Referring to Figure 5, in step 3402, the time delays for these directions are calculated.
参见图5，基于图4得到的所有信号方向的方法中，在步骤3403中，将这些有用信号的时间延迟在所有可能的方向中排除，对剩余的每一个不同的时间延迟分别重复做步骤3103中的同步化处理。再次参见图1，在步骤400中，同步后的输入信号分离为含有目标干扰信号的频道和不含目标干扰信号的频道。优选的，步骤400是通过同步化的信号矩阵和步骤200确定的系数矩阵的乘法运算来实现的。Referring to FIG. 5, in the method for all signal directions obtained based on FIG. 4, in step 3403, the time delay of these useful signals is excluded in all possible directions, and step 3103 is repeated for each of the remaining different time delays. Synchronization in . Referring again to FIG. 1, in step 400, the synchronized input signal is separated into a channel containing the target interference signal and a channel containing no target interference signal. Preferably, step 400 is implemented by a multiplication of the synchronized signal matrix and the coefficient matrix determined in step 200.
举例来说，参见步骤100的例子，假设混合信号组成如下：For example, referring to the example of step 100, assume that the mixed signal is composed as follows:
Figure PCTCN2017117813-appb-000005
Figure PCTCN2017117813-appb-000005
通过步骤200得出的系数矩阵之后，与同步化后的信号矩阵相乘，公式如下：After the coefficient matrix obtained in step 200, multiplied by the synchronized signal matrix, the formula is as follows:
Figure PCTCN2017117813-appb-000006
Figure PCTCN2017117813-appb-000006
由该公式可知，将产生两个频道，其中一个频道
Figure PCTCN2017117813-appb-000007
换言之，该频道由包含96％的S2和4％的S1，如果目标干扰信号为S1的话，该频道就该被选中而输出。因此，在此例，同步化的分离效果达到了96％。According to the formula, two channels will be generated, one of which is
Figure PCTCN2017117813-appb-000007
In other words, the channel is composed of 96% S2 and 4% S1, and if the target interference signal is S1, the channel is selected and output. Therefore, in this case, the separation effect of synchronization is 96%.
如何在此两个频道中选择，会在步骤500中展开。How to choose between these two channels will be expanded in step 500.
同理，若目标干扰信号为S2的话，则将同步S2的混合信号与系数矩阵相乘。同时选择合适的频道输出。Similarly, if the target interference signal is S2, the mixed signal of the synchronization S2 is multiplied by the coefficient matrix. Also select the appropriate channel output.
参见图1，在步骤500中，基于步骤400中得到的两个频道，可以根据不同的信号相对能量来选择，选择信号能量相对较低的那个频道为输出频道。信号能量的计算方法可以是该信号的均方根值。该选择过程运用到从步骤500中得到的含有目标干扰信号的频道和不含目标干扰信号的频道中。Referring to FIG. 1, in step 500, based on the two channels obtained in step 400, the relative energy of the signal can be selected according to different signals, and the channel with the relatively lower signal energy is selected as the output channel. The method of calculating the signal energy can be the root mean square value of the signal. The selection process is applied to the channel containing the target interference signal obtained in step 500 and the channel containing no target interference signal.
进一步地，在图4和5的实施例中，将会在不同的时间延迟中产生一个输出频道，在图4的实施例中，基于特征检测选择最优频道作为信号输出(如产生频道中的目标干扰信号成分最少的频道)；在图5的实施例中可以根据信号能量选择最优频道作为信号输出(如产生频道中目标干扰信号能量最低的频道)。Further, in the embodiments of Figures 4 and 5, an output channel will be generated in different time delays. In the embodiment of Figure 4, the optimal channel is selected as the signal output based on feature detection (e.g., in the generated channel) The channel with the least interference component of the target signal; in the embodiment of Fig. 5, the optimal channel can be selected as the signal output according to the signal energy (e.g., the channel with the lowest energy of the target interference signal in the generated channel).
优选的，在步骤500分离了干扰信号之后，还可以包括对分离后的有用信号和干扰信号进一步处理的步骤，比如做频域增强。例如，在助听器应用中，可以将分离后的有用音频信号做个性化频域增强。Preferably, after the interference signal is separated in step 500, the step of further processing the separated useful signal and the interference signal may be further included, for example, performing frequency domain enhancement. For example, in a hearing aid application, the separated useful audio signal can be personalized for frequency domain enhancement.
在一个实施例里，本公开提供一种装置，其包括处理器，和人工交互界面。该装置还可以包括但不局限于存储器，控制器，输入输出模块， 信息接收模块。处理器用于执行上述步骤100,200,3201-3203(或者3401-3403)，和400，500，及频域增强(可选的)。用户通过人工交互界面实时来选择他希望那个区域为干扰信号区域。人工交互界面包括但不仅限于语音接收模块，传感器，视频接收模块。触控屏，键盘，按钮，旋钮，投影界面，虚拟3D界面。用户通过人工交互界面实时选择的方式包括通过语音指令，通过用户的不同姿势或动作，通过选择不同标识的区域。人工交互界面为触控屏时，用户可以点击其中某个区域，该公开提供一种用户可控可选择的去除干扰信号的机器，并且可以实时地对延迟进行调整。In one embodiment, the present disclosure provides an apparatus that includes a processor, and a human interaction interface. The device may also include, but is not limited to, a memory, a controller, an input and output module, Information receiving module. The processor is configured to perform the above steps 100, 200, 3201-3203 (or 3401-3403), and 400, 500, and frequency domain enhancement (optional). The user selects the area he wants to be the interference signal area in real time through the manual interaction interface. The artificial interaction interface includes but is not limited to a voice receiving module, a sensor, and a video receiving module. Touch screen, keyboard, buttons, knobs, projection interface, virtual 3D interface. The manner in which the user selects in real time through the manual interaction interface includes selecting a different identified area by voice instructions through different gestures or actions of the user. When the human interaction interface is a touch screen, the user can click on one of the areas, and the disclosure provides a user-controllable machine for removing interference signals, and the delay can be adjusted in real time.
上述步骤100-400可能以不同于附图中描述的先后顺序出现。例如，步骤100和步骤300的第二个实施例(即3201-3203)的顺序可以互相调换。再例如，在实际应用中，根据具体涉及的功能，步骤100-400中的任意两个步骤可能平行执行或以相反的顺序执行。The above steps 100-400 may occur in a different order than described in the drawings. For example, the order of the second embodiment of steps 100 and 300 (ie, 3201-3203) can be interchanged. As another example, in practical applications, any two of steps 100-400 may be performed in parallel or in reverse order, depending on the particular functionality involved.
优选的，步骤200在步骤300前实施，即先计算系数矩阵，然后在时域上同步化输入信号。这么做的好处在于，不需要根据不同的时间延迟，都重新计算一次系数矩阵。如此可以大量节约计算量。尤其是在图4和图5所描述的实施例中，只需要计算一次系数矩阵，就可以得到结果。同时，本公开通过进行大量的实验得出结论：同步化之后的混合信号计算出的系数矩阵和原始混合信号计算出的系数矩阵相差无几。因此，这种方法在不损失系数矩阵的精度的同时大量节约了计算量。Preferably, step 200 is performed before step 300 by first calculating a coefficient matrix and then synchronizing the input signal in the time domain. The advantage of this is that there is no need to recalculate the coefficient matrix based on different time delays. This can save a lot of calculations. In particular, in the embodiment depicted in Figures 4 and 5, the result matrix can only be obtained by calculating the coefficient matrix once. At the same time, the present disclosure concludes by conducting a large number of experiments that the coefficient matrix calculated by the mixed signal after synchronization and the coefficient matrix calculated by the original mixed signal are almost the same. Therefore, this method saves a large amount of calculation without losing the accuracy of the coefficient matrix.
优选的，在步骤100中，在使用m个信号接收装置接收n个输入信号之后，根据判断条件决定是否去除多个信号接收装置中一个或多个信号 接收装置所接收的输入信号。Preferably, in step 100, after receiving n input signals by using m signal receiving devices, determining whether to remove one or more signals of the plurality of signal receiving devices according to the determining condition Receiving an input signal received by the device.
在某个实施例里，输入信号是声学信号，信号接收装置为声学信号接收装置(例如麦克风)。判断条件为Fs*X/V<L/3时候(其中L为截取的离散信号长度，X为任意两个声学信号接收装置的距离，V为信号传播速度，Fs为采样率)，将两个声学信号接收装置中的其中一个所接收到的声学信号去除。本实施例在不影响模式识别的精确性的同时减少了需要计算的数据量，提高了计算效率，降低了功耗。In one embodiment, the input signal is an acoustic signal and the signal receiving device is an acoustic signal receiving device (eg, a microphone). When the judgment condition is Fs*X/V<L/3 (where L is the length of the discrete signal intercepted, X is the distance of any two acoustic signal receiving devices, V is the signal propagation speed, Fs is the sampling rate), and two The acoustic signal received by one of the acoustic signal receiving devices is removed. This embodiment reduces the amount of data that needs to be calculated while not affecting the accuracy of pattern recognition, improves computational efficiency, and reduces power consumption.
所述信号包括音频信号、图像信号、电磁信号、脑电波信号、电信号、无线电波信号及其他形式的可被传感器接收的信号，本公开对此并无特定的限制。The signals include audio signals, image signals, electromagnetic signals, brain wave signals, electrical signals, radio wave signals, and other forms of signals that can be received by the sensor, and the disclosure is not particularly limited thereto.
本公开可极大地提升目标信号的感知度同时降低运算成本。此外，本公开输入信号在时域上经过了同步化处理，因此本公开的方法最小化了频率失真。The present disclosure can greatly improve the perception of the target signal while reducing the computational cost. Furthermore, the input signal of the present disclosure is synchronized in the time domain, so the method of the present disclosure minimizes frequency distortion.
图6是本公开的一种适用于实现上述实施方式的计算机***3000的结构示意图。FIG. 6 is a schematic diagram of the structure of a computer system 3000 suitable for implementing the above embodiments.
如图6所示，该计算机***3000包括中央处理器(CPU)3001，该处理器能根据储存在电可编程只读存储器(EPROM)3002或随机存取存储器(RAM)3003上的程序指令执行各种适宜的操作和流程。随机存取存储器(RAM)3003上还可以存储运行所述***3000的必要程序和数据。所述中央处理器(CPU)3001、电可编程只读存储器(EPROM)3002和随机存取存储器(RAM)3003通过总线3004相互连接。所述总线3004还连接有输 入/输出(I/O)接口3005。所述总线3004还连接有直接存储器存取接口3006以便加快数据交换。As shown in FIG. 6, the computer system 3000 includes a central processing unit (CPU) 3001 that is executable in accordance with program instructions stored on an electrically programmable read only memory (EPROM) 3002 or random access memory (RAM) 3003. A variety of suitable operations and processes. The necessary programs and data for running the system 3000 can also be stored on the random access memory (RAM) 3003. The central processing unit (CPU) 3001, the electrically programmable read only memory (EPROM) 3002, and the random access memory (RAM) 3003 are connected to one another via a bus 3004. The bus 3004 is also connected to lose In/Out (I/O) interface 3005. The bus 3004 is also coupled to a direct memory access interface 3006 to facilitate data exchange.
所述输入/输出(I/O)接口3005还连接有下列元件：可移动数据存储器3007，包括USB存储器，固态硬盘等；无线数据传输线路3008，包括局域网(LAN),蓝牙，近场通信设备(NFC)；以及与数据输入通道3010和数据输出通道3011相连接的信号变换器3009。根据本公开的另一实施例，上述流程图涉及的过程可由与所述计算机***3000相似的不带键盘、鼠标和硬盘的嵌入式计算机***实现。无线数据传输线路3008或可移动数据存储器3007有利于程序的更新升级。The input/output (I/O) interface 3005 is also connected to the following components: a removable data storage 3007, including a USB memory, a solid state hard disk, etc.; a wireless data transmission line 3008, including a local area network (LAN), a Bluetooth, a near field communication device (NFC); and a signal converter 3009 connected to the data input channel 3010 and the data output channel 3011. According to another embodiment of the present disclosure, the process of the above flow chart may be implemented by an embedded computer system without a keyboard, mouse, and hard disk similar to the computer system 3000. The wireless data transmission line 3008 or the removable data storage 3007 facilitates the update of the program.
其中，上述处理器可以为云端处理器，存储器可以为云端存储器。The processor may be a cloud processor, and the memory may be a cloud memory.
进一步地，根据本公开的又一实施例，上述流程图涉及的过程可由计算机软件程序实现。例如，本公开实施例提供一种计算机程序产品，其包括存储于有形的机读介质中的计算机程序，该程序包括执行流程图中所示方法的程序代码。在本实施例中，所述计算机程序可通过无线数据传输线路3008下载和安装，以及/或者从可移动介质3007安装。Further, according to still another embodiment of the present disclosure, the process involved in the above flowchart may be implemented by a computer software program. For example, embodiments of the present disclosure provide a computer program product comprising a computer program stored in a tangible machine readable medium, the program comprising program code for executing the method shown in the flowchart. In the present embodiment, the computer program can be downloaded and installed via wireless data transmission line 3008 and/or installed from removable medium 3007.
附图中的流程图和方框图阐释了本公开不同实施例中***、方法及计算机程序产品的构架、功能及操作的实现过程。就此而言，流程图和方框图中的每一个方框代表一种模块、程序段或代码单元。所述模块、程序段或代码单元包含一个或多个用于实现指定的逻辑功能的可执行指令。需要指出的是，在某些优选实施例中，模块表示的功能可能以不同于附图中描述的先后顺序出现。例如，在实际应用中，根据具体涉及的功能，相连 的两个方框所示的操作可能平行执行或以相反的顺序执行。同时需要说明的是，流程图和/或方框图中的每一个方框或其组合可通过一个专用的、以硬件为基础的、可执行特定功能或运算的体系实现，或通过专用软件和计算机指令的组合实现。The flowchart and block diagrams in the Figures illustrate the implementation of the architecture, functionality, and operation of systems, methods, and computer program products in various embodiments of the present disclosure. In this regard, each block in the flowcharts and block diagrams represents a module, a program segment, or a code unit. The module, program segment or code unit contains one or more executable instructions for implementing the specified logical function. It is pointed out that in certain preferred embodiments, the functions represented by the modules may occur in a different order than that described in the drawings. For example, in practical applications, connected according to the specific functions involved. The operations shown in the two blocks may be performed in parallel or in reverse order. It should also be noted that each block or combination of the flowcharts and/or block diagrams can be implemented by a dedicated, hardware-based system that can perform a particular function or operation, or through dedicated software and computer instructions. The combination is implemented.
本公开实施例中涉及的单元或模块可通过软件或硬件实现，所述单元或模块可安装在处理器中。所述单元或模块的名称不应该对单元或模块的自身造成限定。The unit or module involved in the embodiments of the present disclosure may be implemented by software or hardware, and the unit or module may be installed in a processor. The name of the unit or module should not limit the unit or module itself.
另一方面，本公开还提供一种计算机可读存储介质。所述计算机可读存储介质可以是安装在上述实施例应用的仪器设备中的计算机可读存储介质，也可以是独立的并不装配于仪器设备中的计算机可读存储介质。所述计算机可读存储介质存储一个或多个程序，所述程序由一个或多个处理器执行以实现本公开的从噪音信号中分选目标信号的方法。In another aspect, the present disclosure also provides a computer readable storage medium. The computer readable storage medium may be a computer readable storage medium installed in an instrument device to which the above embodiments are applied, or may be a separate computer readable storage medium that is not assembled in an instrument device. The computer readable storage medium stores one or more programs that are executed by one or more processors to implement the method of sorting a target signal from a noise signal of the present disclosure.
以上内容是结合具体的优选实施方式对本公开所作的进一步详细说明，不能认定本公开的具体实施只局限于这些说明。对于本公开所属技术领域的普通技术人员来说，在不脱离本公开构思的前提下，还可以做出若干简单推演或替换，都应当视为属于本公开的保护范围。 The above is a further detailed description of the present disclosure in conjunction with the specific preferred embodiments, and the specific embodiments of the present disclosure are not limited to the description. It is to be understood by those skilled in the art that the present invention may be construed as being limited to the scope of the present disclosure.

Claims (15)

  1. 一种从多重信号中去除目标干扰信号的方法，其特征在于，该方法包括：A method for removing a target interference signal from a multiple signal, the method comprising:
    接收一组输入信号，该组输入信号中的每条输入信号既包含有用信号也包含干扰信号；Receiving a set of input signals, each of the input signals comprising both a useful signal and an interference signal;
    提高所述输入信号的独立性；Increasing the independence of the input signal;
    计算提升所述输入信号的独立性所得的系数矩阵；Calculating a coefficient matrix obtained by increasing the independence of the input signal;
    同步所述输入信号；Synchronizing the input signal;
    将同步后的输入信号分离为含有目标干扰信号的频道和不含目标干扰信号的频道；Separating the synchronized input signal into a channel containing the target interference signal and a channel containing no target interference signal;
    智能选择不含目标干扰信号的频道作为信号输出。Intelligently select the channel without the target interference signal as the signal output.
  2. 根据权利要求1所述的方法，其特征在于，所述同步所述输入信号的操作包括：The method of claim 1 wherein said synchronizing said input signal comprises:
    检测每条输入信号中的干扰信号片段；Detecting interfering signal segments in each input signal;
    对检测到的每两个干扰信号片段进行离散时间卷积运算以获取相对时间延迟；Performing a discrete time convolution operation on each of the detected two interference signal segments to obtain a relative time delay;
    基于所获取的时间延迟，同步所述输入信号；Synchronizing the input signal based on the acquired time delay;
    选择被标记为干扰信号的信号优先方向；Select the priority direction of the signal marked as an interfering signal;
    计算来自优先方向的干扰信号的相对时间延迟；Calculating the relative time delay of the interfering signal from the preferential direction;
    基于预设的时间延迟，同步所述输入信号；Synchronizing the input signal based on a preset time delay;
    选择被标记为干扰信号的所有可能的信号方向；Select all possible signal directions that are marked as interfering signals;
    预估一系列时间延迟，记为τ1,τ2,…,τn； Estimate a series of time delays, denoted as τ1, τ2,...,τn;
    基于一系列的时间延迟，同步所述输入信号；Synchronizing the input signal based on a series of time delays;
    选择被标记为有用信号的信号进入方向；Select the direction of the signal marked as a useful signal;
    确定来自其余方向的干扰信号的时间延迟；Determining the time delay of the interfering signal from the remaining directions;
    基于确定的时间延迟，同步所述干扰信号。The interference signal is synchronized based on the determined time delay.
  3. 根据权利要求1或2所述的方法，其特征在于，所述输入信号的同步化可持续升级以便适应信号源的运动状态。Method according to claim 1 or 2, characterized in that the synchronization of the input signals is continuously upgraded in order to adapt to the state of motion of the signal source.
  4. 根据权利要求1-3任一条所述的方法，其特征在于，所述输入信号取自于相互间隔的位点。A method according to any of claims 1-3, wherein said input signal is taken from mutually spaced sites.
  5. 根据权利要求1-4任一条所述的方法，其特征在于，所述提高所述输入信号的独立性包括：通过独立成分分析最大化所述输入信号的高斯分布。A method according to any one of claims 1 to 4, wherein said increasing the independence of said input signal comprises maximizing a Gaussian distribution of said input signal by independent component analysis.
  6. 根据权利要求2所述的方法，其特征在于，所述检测每条输入信号中的干扰信号片段包括：通过模式识别检测每条输入信号中的干扰信号片段。The method of claim 2 wherein said detecting the interfering signal segments in each of the input signals comprises detecting the interfering signal segments in each of the input signals by pattern recognition.
  7. 根据权利要求1所述的方法，其特征在于，所述输入信号是由传感器接收的信号。The method of claim 1 wherein said input signal is a signal received by a sensor.
  8. 根据权利要求1-7任一条所述的方法，其特征在于，所述输入信号是以下中的一种：A method according to any of claims 1-7, wherein said input signal is one of the following:
    音频信号；audio signal;
    电信号；electric signal;
    图像信号；及 Image signal; and
    射频信号。RF signal.
  9. 一种从信号中去除目标噪音的***，其特征在于，其包括：A system for removing target noise from a signal, characterized in that it comprises:
    一套用于输入一组输入信号的输入设备；a set of input devices for inputting a set of input signals;
    处理器；以及Processor;
    存储计算机可读指令的存储器，当所述处理器执行所述指令时，该处理器可进行：A memory storing computer readable instructions, when the processor executes the instructions, the processor can:
    提升所述输入信号的独立性；Enhancing the independence of the input signal;
    计算在输入通道提升所述输入信号的独立性所得的系数矩阵；Calculating a coefficient matrix obtained by increasing the independence of the input signal in the input channel;
    同步所述输入信号；Synchronizing the input signal;
    将同步后的输入信号分离为含有目标干扰信号的频道和不含目标干扰信号的频道；Separating the synchronized input signal into a channel containing the target interference signal and a channel containing no target interference signal;
    智能选择不含目标干扰信号的频道作为信号输出。Intelligently select the channel without the target interference signal as the signal output.
  10. 根据权利要求9所述的***，其特征在于，所述同步所述输入信号包括：The system of claim 9 wherein said synchronizing said input signal comprises:
    检测每条输入信号中的干扰信号片段；Detecting interfering signal segments in each input signal;
    对检测到的每两个干扰信号片段进行离散时间卷积运算以获取相对时间延迟；Performing a discrete time convolution operation on each of the detected two interference signal segments to obtain a relative time delay;
    基于所获取的时间延迟，同步所述输入信号；Synchronizing the input signal based on the acquired time delay;
    选择被标记为干扰信号的信号优先方向；Select the priority direction of the signal marked as an interfering signal;
    计算来自优先方向的干扰信号的相对时间延迟；Calculating the relative time delay of the interfering signal from the preferential direction;
    基于预设的时间延迟，同步所述输入信号； Synchronizing the input signal based on a preset time delay;
    选择被标记为干扰信号的所有可能的信号方向；Select all possible signal directions that are marked as interfering signals;
    预估一系列时间延迟，记为τ1,τ2,…,τn；Estimate a series of time delays, denoted as τ1, τ2,...,τn;
    基于一系列的时间延迟，同步所述输入信号；Synchronizing the input signal based on a series of time delays;
    选择被标记为有用信号的信号进入方向；Select the direction of the signal marked as a useful signal;
    确定来自其余方向的干扰信号的时间延迟；Determining the time delay of the interfering signal from the remaining directions;
    基于确定的时间延迟，同步所述干扰信号。The interference signal is synchronized based on the determined time delay.
  11. 根据权利要求9或10所述的***，其特征在于，所述输入信号取自于相互间隔的位点。The system of claim 9 or 10 wherein said input signals are taken from mutually spaced locations.
  12. 根据权利要求9或10所述的***，其特征在于，所述提升所述输入信号的独立性包括：通过独立成分分析最大化所述输入信号的高斯分布。The system of claim 9 or 10, wherein said increasing the independence of said input signal comprises maximizing a Gaussian distribution of said input signal by independent component analysis.
  13. 根据权利要求10所述的***，其特征在于，所述检测每条输入信号中的干扰信号片段包括：通过模式识别检测每条输入信号中的干扰信号片段。The system of claim 10 wherein said detecting the interfering signal segments in each of the input signals comprises detecting the interfering signal segments in each of the input signals by pattern recognition.
  14. 根据权利要求9或10所述的***，其特征在于，所述输入信号是由传感器接收的信号，为以下中的一种：A system according to claim 9 or 10, wherein said input signal is a signal received by a sensor and is one of:
    音频信号；audio signal;
    电信号；electric signal;
    图像信号；Image signal
    射频信号。RF signal.
  15. 一种非临时性计算机可读存储介质，其特征在于，该存储介质存储有指令，当处理器执行该指令时，可实现一种从多重信号中分离目标干 扰信号的方法，所述方法包括：A non-transitory computer readable storage medium, characterized in that the storage medium stores instructions that, when executed by the processor, implement a separation of the target from the multiple signals A method of scrambling a signal, the method comprising:
    接收一组输入信号(观测信号)，每条所述输入信号均含有目标干扰信号；Receiving a set of input signals (observation signals), each of the input signals containing a target interference signal;
    提升所述输入信号的独立性；Enhancing the independence of the input signal;
    计算提升所述输入信号的独立性所得的系数矩阵；Calculating a coefficient matrix obtained by increasing the independence of the input signal;
    同步所述输入信号；Synchronizing the input signal;
    将同步后的输入信号分离为含有目标干扰信号的频道和不含目标干扰信号的频道；Separating the synchronized input signal into a channel containing the target interference signal and a channel containing no target interference signal;
    智能选择不含目标干扰信号的频道作为信号输出。 Intelligently select the channel without the target interference signal as the signal output.
PCT/CN2017/117813 2017-08-15 2017-12-21 Method and system for extracting source signal, and storage medium WO2019033671A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP17921701.3A EP3672275A4 (en) 2017-08-15 2017-12-21 Method and system for extracting source signal, and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710698651.6 2017-08-15
CN201710698651.6A CN109413543B (en) 2017-08-15 2017-08-15 Source signal extraction method, system and storage medium

Publications (1)

Publication Number Publication Date
WO2019033671A1 true WO2019033671A1 (en) 2019-02-21

Family

ID=65362112

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/117813 WO2019033671A1 (en) 2017-08-15 2017-12-21 Method and system for extracting source signal, and storage medium

Country Status (3)

Country Link
EP (1) EP3672275A4 (en)
CN (1) CN109413543B (en)
WO (1) WO2019033671A1 (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1812312A (en) * 2006-02-20 2006-08-02 东南大学 Space time separation soft inputting and outputting detecting method based on spatial domain prewhitening mergence
CN101162453A (en) * 2006-10-10 2008-04-16 深圳市理邦精密仪器有限公司 Signal processing method using space coordinates convert realizing signal separation
US20120099732A1 (en) * 2010-10-22 2012-04-26 Qualcomm Incorporated Systems, methods, apparatus, and computer-readable media for far-field multi-source tracking and separation
CN102868433A (en) * 2012-09-10 2013-01-09 西安电子科技大学 Signal transmission method based on antenna selection in multiple-input multiple-output Y channel
CN103197183A (en) * 2013-01-11 2013-07-10 北京航空航天大学 Method for correcting uncertainty of independent component analytical method in electromagnetic interference separation
US20130272445A1 (en) * 2010-12-07 2013-10-17 Huawei Technologies Co., Ltd. Precoding method and apparatus

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2430975A1 (en) * 2010-09-17 2012-03-21 Stichting IMEC Nederland Principal component analysis or independent component analysis applied to ambulatory electrocardiogram signals
JP2012234150A (en) * 2011-04-18 2012-11-29 Sony Corp Sound signal processing device, sound signal processing method and program
US9099096B2 (en) * 2012-05-04 2015-08-04 Sony Computer Entertainment Inc. Source separation by independent component analysis with moving constraint
US8880395B2 (en) * 2012-05-04 2014-11-04 Sony Computer Entertainment Inc. Source separation by independent component analysis in conjunction with source direction information
JP2014045793A (en) * 2012-08-29 2014-03-17 Sony Corp Signal processing system, signal processing apparatus, and program
CN103083012A (en) * 2012-12-24 2013-05-08 太原理工大学 Atrial fibrillation signal extraction method based on blind source separation
CN104053107B (en) * 2014-06-06 2018-06-05 重庆大学 One kind is for Sound seperation and localization method under noise circumstance
CN104091356A (en) * 2014-07-04 2014-10-08 南京邮电大学 X-ray medical image objective reconstruction based on independent component analysis
CN108353228B (en) * 2015-11-19 2021-04-16 香港科技大学 Signal separation method, system and storage medium
CN105640500A (en) * 2015-12-21 2016-06-08 安徽大学 Independent component analysis-based saccade signal feature extraction method and recognition method
CN105996993A (en) * 2016-04-29 2016-10-12 南京理工大学 System and method for intelligent video monitoring of vital signs
CN106356075B (en) * 2016-09-29 2019-09-17 合肥美的智能科技有限公司 Blind sound separation method, structure and speech control system and electric appliance assembly
CN107025446A (en) * 2017-04-12 2017-08-08 北京信息科技大学 A kind of vibration signal combines noise-reduction method

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1812312A (en) * 2006-02-20 2006-08-02 东南大学 Space time separation soft inputting and outputting detecting method based on spatial domain prewhitening mergence
CN101162453A (en) * 2006-10-10 2008-04-16 深圳市理邦精密仪器有限公司 Signal processing method using space coordinates convert realizing signal separation
US20120099732A1 (en) * 2010-10-22 2012-04-26 Qualcomm Incorporated Systems, methods, apparatus, and computer-readable media for far-field multi-source tracking and separation
US20130272445A1 (en) * 2010-12-07 2013-10-17 Huawei Technologies Co., Ltd. Precoding method and apparatus
CN102868433A (en) * 2012-09-10 2013-01-09 西安电子科技大学 Signal transmission method based on antenna selection in multiple-input multiple-output Y channel
CN103197183A (en) * 2013-01-11 2013-07-10 北京航空航天大学 Method for correcting uncertainty of independent component analytical method in electromagnetic interference separation

Also Published As

Publication number Publication date
CN109413543A (en) 2019-03-01
CN109413543B (en) 2021-01-19
EP3672275A4 (en) 2023-08-23
EP3672275A1 (en) 2020-06-24

Similar Documents

Publication Publication Date Title
US9668066B1 (en) Blind source separation systems
CN111133511B (en) sound source separation system
Mandel et al. An EM algorithm for localizing multiple sound sources in reverberant environments
US20200089468A1 (en) Audio synchronization and delay estimation
US20120082322A1 (en) Sound scene manipulation
GB2548325A (en) Acoustic source seperation systems
JP2007033825A (en) Device, program, and method for sound source separation
WO2015065682A1 (en) Selective audio source enhancement
JPWO2015159731A1 (en) Sound field reproduction apparatus and method, and program
JP7214798B2 (en) AUDIO SIGNAL PROCESSING METHOD, AUDIO SIGNAL PROCESSING DEVICE, ELECTRONIC DEVICE, AND STORAGE MEDIUM
Talagala et al. Binaural sound source localization using the frequency diversity of the head-related transfer function
WO2022256577A1 (en) A method of speech enhancement and a mobile computing device implementing the method
CN108353228B (en) Signal separation method, system and storage medium
CN113168843B (en) Audio processing method and device, storage medium and electronic equipment
Hosseini et al. Time difference of arrival estimation of sound source using cross correlation and modified maximum likelihood weighting function
WO2019033671A1 (en) Method and system for extracting source signal, and storage medium
US11823698B2 (en) Audio cropping
Jafari et al. Underdetermined blind source separation with fuzzy clustering for arbitrarily arranged sensors
JP2010217268A (en) Low delay signal processor generating signal for both ears enabling perception of direction of sound source
JP2003078423A (en) Processor for separating blind signal
US10887697B2 (en) Method, system and apparatus for extracting target unwanted audio signal from mixture of audio signals
CN115910047B (en) Data processing method, model training method, keyword detection method and equipment
JP2006072163A (en) Disturbing sound suppressing device
CN109951762B (en) Method, system and device for extracting source signal of hearing device
JP2011071683A (en) Video object detection apparatus, video object detection method and program

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17921701

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2017921701

Country of ref document: EP

Effective date: 20200316